Battery capacity calculation method of storage and charging system

Abstract

The invention discloses a storage and charging system battery capacity calculation method, and relates to the technical field of storage and charging system optimization control, and the method comprises the steps: collecting multi-modal data of a battery pack, and generating time-space aligned multi-modal data through a timestamp alignment and three-dimensional coordinate conversion algorithm; extracting local temperature gradient characteristics of an infrared thermogram and dynamic time characteristics of a charging and discharging curve through DSC and Bi LSTM based on the multi-modal data of space-time alignment, segmenting spatial distribution characteristics of electrolyte flow abnormity of an ultrasonic image by using U-Net, dynamically distributing weights of the multi-modal characteristics through a space-time attention mechanism, and generating a fusion characteristic vector; and inputting the fusion feature vector into a GRU-Transform hybrid model to generate a dynamic capacity prediction value, and identifying an aging type by dynamically analyzing historical data. According to the invention, multi-modal features are extracted through DSC, Bi LSTM, U-Net and other technologies, and comprehensive monitoring and dynamic adjustment of the battery health state are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of optimized control of energy storage and charging systems, and particularly to a method for calculating the battery capacity of an energy storage and charging system. Background Art

[0002] With the rapid development of renewable energy technologies, the importance of energy storage systems in power grid peak shaving, distributed energy management, and electric vehicles has become increasingly prominent. As the core component of an energy storage system, the performance and lifespan of the battery directly affect the reliability and economy of the entire system. In recent years, to more accurately evaluate the state of the battery and predict its capacity changes, researchers have developed various methods based on sensor data and machine learning algorithms.

[0003] Although existing BMSs have achieved a certain degree of monitoring of the battery health state, there are still several deficiencies. Firstly, traditional methods usually rely only on electrochemical parameters for analysis, ignoring other important information sources, resulting in an incomplete understanding of the internal state of the battery. Secondly, most existing methods use fixed model parameters and lack an adaptive adjustment mechanism, unable to dynamically respond to the complex changes during the battery aging process. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method for calculating the battery capacity of an energy storage and charging system to solve the problems of single data source and fixed model parameters in the prior art.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for calculating the battery capacity of an energy storage and charging system, which includes collecting multimodal data of a battery pack, and generating spatio-temporally aligned multimodal data through a timestamp alignment and three-dimensional coordinate transformation algorithm; based on the spatio-temporally aligned multimodal data, extracting local temperature gradient features of an infrared thermal image and dynamic time features of a charge-discharge curve through DSC and BiLSTM, using U-Net to segment the spatial distribution features of abnormal electrolyte flow in an ultrasonic image, and dynamically allocating weights to multimodal features through a spatio-temporal attention mechanism to generate a fused feature vector; inputting the fused feature vector into a GRU-Transformer hybrid model to generate a dynamic capacity prediction value, and identifying the aging type by dynamically analyzing historical data; based on the aging type, displaying the lithium-ion distribution through the color change of a photochromic electrolyte, calculating the local impedance value of the lithium dendrite growth region using the principle of quantum tunneling, calibrating the battery interface state, automatically updating the capacity prediction model parameters, and outputting a calibrated capacity value.

[0008] As a preferred solution of the battery capacity calculation method for the storage and charging system of the present invention, wherein: the multimodal data includes infrared thermal imaging temperature field data, ultrasonic electrolyte distribution data, visible light appearance deformation image data, and electrochemical parameter data.

[0009] As a preferred solution of the battery capacity calculation method for the storage and charging system of the present invention, wherein: through the timestamp alignment and three-dimensional coordinate conversion algorithm, generate spatio-temporally aligned multimodal data, and the specific steps are as follows.

[0010] Add synchronous timestamps to the multimodal data, and compensate for the sampling rate difference through the cubic spline interpolation algorithm to generate time-synchronized data values.

[0011] Based on the three-dimensional structure model of the battery pack, map the time-synchronized data values to the same coordinate system, eliminate the position deviation through the rotation and translation matrix, and output the spatio-temporally aligned multimodal data.

[0012] As a preferred solution of the battery capacity calculation method for the storage and charging system of the present invention, wherein: based on the spatio-temporally aligned multimodal data, extract the local temperature gradient features of the infrared thermal image and the dynamic time features of the charge and discharge curve through DSC and BiLSTM, and use U-Net to segment the spatial distribution features of the abnormal electrolyte flow in the ultrasonic image. The specific steps are as follows.

[0013] Input the infrared thermal image in the spatio-temporally aligned multimodal data into DSC, and extract the local temperature gradient features layer by layer through the convolutional kernel.

[0014] Input the charge and discharge voltage curve in the spatio-temporally aligned multimodal data into BiLSTM, and extract the dynamic time features of the charge and discharge curve through the bidirectional sequence processing ability of BiLSTM.

[0015] Input the ultrasonic image in the spatio-temporally aligned multimodal data into the U-Net segmentation network, and generate the spatial distribution features of the abnormal electrolyte flow through the encoder-decoder structure.

[0016] As a preferred solution of the battery capacity calculation method for the storage and charging system of the present invention, wherein: dynamically allocate the weights of the multimodal features through the spatio-temporal attention mechanism to generate a fused feature vector. The specific steps are as follows.

[0017] Input the local temperature gradient features of the infrared thermal image, the dynamic time features of the charge and discharge curve, and the spatial distribution features of the abnormal electrolyte flow in the spatio-temporally aligned multimodal data into the fully connected layer to generate corresponding feature encoding vectors respectively.

[0018] Based on the feature encoding vectors, dynamically allocate the weight features of the multimodal through the spatio-temporal attention mechanism.

[0019] Dynamically adjust the weight of the local temperature gradient feature of the infrared thermal image according to the maximum value of the local temperature gradient;

[0020] Dynamically adjust the weight of the dynamic time feature of the charge-discharge curve according to the slope of the charge-discharge curve;

[0021] Dynamically adjust the weight of the spatial distribution feature of abnormal electrolyte flow according to the proportion of the area of the abnormal electrolyte region;

[0022] Generate a fused feature vector by concatenating the local temperature gradient feature of the infrared thermal image, the dynamic time feature of the charge-discharge curve, and the spatial distribution feature of abnormal electrolyte flow along the dimension.

[0023] As a preferred solution of the battery capacity calculation method for the storage and charging system described in the present invention, wherein: input the fused feature vector into the GRU-Transformer hybrid model, output the dynamic capacity prediction value after fusion, and identify the aging type by dynamically analyzing historical data. The specific steps are as follows.

[0024] Input the fused feature vector into the bidirectional GRU unit in the order of time windows, extract the voltage-temperature-internal resistance time series correlation features under charge-discharge cycles, and output a sequence of hidden state vectors;

[0025] Input the sequence of hidden state vectors into the Transformer encoder, identify the aging correlation matrix between features at different time steps through the self-attention mechanism, and generate an encoded feature vector;

[0026] Fuse the encoded feature vector with the state of charge of the real-time battery, and calculate the dynamic capacity prediction value through a fully connected layer;

[0027] Based on the dynamic capacity prediction value, analyze by extracting the gradient feature of the dynamic capacity prediction value and the historical aging pattern library, and identify the aging type.

[0028] As a preferred solution of the battery capacity calculation method for the storage and charging system described in the present invention, wherein: based on the aging type, display the lithium ion distribution through the color change of the photochromic electrolyte, and calculate the local impedance value of the lithium dendrite growth region using the principle of quantum tunneling. The specific steps are as follows.

[0029] Based on the aging type, send a pulsed voltage signal proportional to the aging degree to the electrolyte injection unit, and activate the color change reaction of the photosensitive compound in the electrolyte;

[0030] Based on the electrolyte color change reaction, capture the color intensity distribution map of the lithium ion deposition region through the multispectral imaging unit, and mark the color mutation coordinates;

[0031] According to the color mutation coordinates, control the movement of the nano-probe array to the target area, apply a constant bias voltage, and start detecting the amplitude change of the current by quantum tunneling current;

[0032] Based on the amplitude change of the quantum tunneling current and the probe displacement, calculate the local impedance value of the lithium dendrite growth area.

[0033] As a preferred solution of the battery capacity calculation method of the charge and discharge system described in the present invention, wherein: after calibrating the battery interface state, automatically update the capacity prediction model parameters and output the calibrated capacity value. The specific steps are as follows.

[0034] Based on the local impedance value, analyze the local impedance change of the lithium dendrite growth area through electrochemical impedance spectroscopy;

[0035] Based on the historical battery charge and discharge data and the interface health status marker, construct a battery interface state evaluation framework, and input the local impedance change into the evaluation framework to generate the corrected interface state parameters;

[0036] Based on the corrected interface state parameters, update the GRU-Transformer hybrid model parameters and output the calibrated capacity value.

[0037] In a second aspect, the present invention provides a computer device, including a memory and a processor. The memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the battery capacity calculation method of the charge and discharge system described in the first aspect of the present invention is implemented.

[0038] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, wherein: when the computer program is executed by the processor, any step of the battery capacity calculation method of the charge and discharge system described in the first aspect of the present invention is implemented.

[0039] The beneficial effects of the present invention are as follows: ensuring the spatio-temporal consistency of data through the timestamp alignment and three-dimensional coordinate conversion algorithm, solving the problem of single data source in traditional methods; and extracting multi-modal features through technologies such as DSC, BiLSTM, and U-Net, realizing the comprehensive monitoring and dynamic adjustment of the battery health state. Description of the Drawings

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0041] Figure 1Flow chart for collecting multi-modal data of the battery pack and spatio-temporal alignment in the battery capacity calculation method of the energy storage and charging system in Embodiment 1.

[0042] Figure 2 Schematic diagram for multi-modal feature extraction in the battery capacity calculation method of the energy storage and charging system in Embodiment 1.

[0043] Figure 3 Flow chart for capacity prediction using a hybrid model in the battery capacity calculation method of the energy storage and charging system in Embodiment 1.

[0044] Figure 4 Flow chart for calibrating the battery interface state and updating the capacity prediction model parameters in the battery capacity calculation method of the energy storage and charging system in Embodiment 1. Detailed implementation manners

[0045] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the detailed implementation manners of the present invention with reference to the accompanying drawings of the specification.

[0046] In the following description, many specific details are set forth to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0047] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selectively exclusive embodiment from other embodiments.

[0048] Embodiment 1, referring to Figures 1 to 4 , which is the first embodiment of the present invention. This embodiment provides a battery capacity calculation method for an energy storage and charging system, including the following steps:

[0049] S1: Collect multi-modal data of the battery pack, and generate spatio-temporally aligned multi-modal data through the time stamp alignment and three-dimensional coordinate conversion algorithm.

[0050] S1.1: The multi-modal data includes infrared thermal imaging temperature field data, ultrasonic electrolyte distribution data, visible light appearance deformation image data, and electrochemical parameter data.

[0051] It should be noted that the acquisition of multimodal data of the battery pack is achieved by integrating multiple sensor technologies, including an infrared thermal imager, an ultrasonic detector, a visible light camera, and an electrochemical sensor. Specifically, the infrared thermal imager is used to obtain the temperature field distribution data on the battery surface, the ultrasonic detector is used to detect the flow and distribution of the electrolyte, the visible light camera is used to capture the deformation images of the battery appearance, and the electrochemical sensor is used to monitor the electrochemical parameters such as the voltage, current, and internal resistance of the battery.

[0052] S1.2: Add synchronous timestamps to the multimodal data and compensate for the sampling rate difference through the cubic spline interpolation algorithm to generate time-synchronized data values. The expression is:

[0053] f(t) = a i (t - t i ) 3 + b i (t - t i ) 2 + c i (t - t i ) + d i (t i ≤ t ≤ t i+1 );

[0054] Among them, f(t) represents the synchronized data value calculated by the cubic polynomial at time t, t represents the time, i represents the segmentation area number of the interpolation function, a i represents the coefficient of the cubic polynomial, t i represents the left endpoint of the i-th time interval, t i+1 represents the right endpoint of the i-th time interval, b i represents the quadratic term coefficient, c i represents the linear term coefficient, d i represents the constant term, a i (t - t i ) 3 represents the non-linear mutation amplitude of lithium deposition regulated by the cubic term coefficient a i within this interval, b i (t - t i ) 2 represents the cumulative effect of the lithium ion diffusion resistance caused by the electrolyte concentration gradient characterized by the quadratic term coefficient b i , c i (t - t i ) represents the charge transfer impedance increment caused by the local crack propagation of the SEI film quantified by the linear term coefficient c i .

[0055] Furthermore, add a unified time stamp to each data record obtained from different sensors to ensure their correspondence in the time dimension. For data with different sampling rates, use the cubic spline interpolation method to perform interpolation within each time interval. Cubic spline interpolation uses a cubic polynomial function to smoothly transition between known data points, generating additional data points to fill the gaps in the sampling intervals. Adjust the data from different sensors to the same sampling rate to generate a set of data values that are completely synchronized in time, ensuring that all data can be comprehensively analyzed under the same time reference.

[0056] S1.3: Based on the three-dimensional structure model of the battery pack, map the time-synchronized data values to the same coordinate system, eliminate the position deviation through the rotation and translation matrix, and output the spatio-temporally aligned multimodal data.

[0057] It should be noted that using the three-dimensional structure model of the battery pack as a reference framework, convert the time-synchronized data values collected by all sensors (such as infrared thermal imaging temperature field data, ultrasonic electrolyte distribution data, etc.) into this unified spatial coordinate system. Specifically, each type of data carries its original spatial position information, and through mathematical transformation, it is mapped into the coordinate system of the three-dimensional model of the battery pack.

[0058] Furthermore, to eliminate the position deviation between the data of different sensors, a rotation and translation matrix is used for correction. The rotation and translation matrix is a linear transformation tool that can simultaneously adjust the position and direction of data points. For each data point of a sensor, calculate the corresponding rotation and translation parameters according to its actual position relative to the three-dimensional structure model of the battery pack. These parameters are integrated into a matrix, and applying this matrix can accurately align the data points to the correct spatial positions. For example, if the position of a data point of a certain sensor in the actual space does not match the expectation, it can be moved to the correct position through appropriate rotation and translation operations.

[0059] S2: Based on the spatio-temporally aligned multimodal data, extract the local temperature gradient features of the infrared thermal image and the dynamic time features of the charge and discharge curve through DSC and BiLSTM, use U-Net to segment the spatial distribution features of the electrolyte flow anomaly in the ultrasonic image, and dynamically allocate the weights of the multimodal features through the spatio-temporal attention mechanism to generate a fused feature vector.

[0060] S2.1: Input the infrared thermal image in the spatio-temporally aligned multimodal data into DSC, and extract the local temperature gradient features layer by layer through the convolutional kernel.

[0061] It should be noted that the infrared thermal image is fed as an input into the first layer of the DSC network. In this layer, multiple convolutional kernels are used to scan the image. Each convolutional kernel performs a convolution operation within the local area it covers, thereby generating a series of feature maps. These convolutional kernels can detect temperature change patterns within the local area, such as temperature gradients and edge information.

[0062] Furthermore, the convolutional kernel slides through the entire infrared thermal image, performs a dot product operation with the corresponding local area at each position, and sums up the results into a new value to form a pixel point on a feature map. After multiple convolutional and pooling operations, the DSC network can extract rich local temperature gradient features from the infrared thermal image, which are crucial for subsequent analysis of the internal state of the battery and identification of potential problems (such as hot spots or abnormal temperature rises).

[0063] S2.2: Input the charge-discharge voltage curve in the spatio-temporally aligned multi-modal data into the BiLSTM, and extract the dynamic time features of the charge-discharge curve through the bidirectional sequence processing ability of the BiLSTM.

[0064] Furthermore, the charge-discharge voltage curve, as time series data, is segmented into multiple consecutive time steps and input into the BiLSTM network. The uniqueness of the BiLSTM lies in that it can not only process time series data from front to back (forward propagation), but also process the same data from back to front (backward propagation). At each time step, the BiLSTM unit updates its internal state based on the current input (i.e., the voltage value at a specific time point) and the state of the previous time step.

[0065] It should be noted that when the data of the charge-discharge voltage curve enters the BiLSTM in sequence, the forward LSTM layer starts to learn and remember the trends and patterns that gradually emerge over time, while the backward LSTM layer complements the influence of future information on the current moment. The hidden states of both are merged at each time point to form a comprehensive representation that contains all relevant information before and after that time point.

[0066] S2.3: Input the ultrasonic image in the spatio-temporally aligned multi-modal data into the U-Net segmentation network, and generate the spatial distribution features of abnormal electrolyte flow through the encoder-decoder structure.

[0067] Further, the ultrasonic image is fed as input into the encoder part of the U-Net network. The encoder consists of a series of convolutional layers and pooling layers, which gradually reduce the spatial resolution of the image while increasing the depth of the feature map to extract high-level abstract features. Specifically, each convolutional operation uses multiple filters to scan the input image and extract patterns and features within local regions. Subsequently, operations such as max pooling or average pooling are used to reduce the size of the feature map, retaining the most important information. As the number of layers deepens, the encoder can capture more complex structures and patterns in the ultrasonic image, such as the direction of electrolyte flow, velocity changes, and potential abnormal regions.

[0068] It should be noted that the high-level feature maps generated by the encoder are passed to the decoder part. The decoder also consists of a series of convolutional layers and upsampling layers, but its role is to gradually restore the spatial resolution of the image while maintaining the important features obtained from the encoder. In each decoder layer, the upsampling operation (such as transposed convolution or bilinear interpolation) increases the size of the feature map, and then it is concatenated with the feature map from the corresponding layer of the encoder, thereby fusing high and low-level information.

[0069] S2.4: Input the local temperature gradient features of the infrared thermal image, the dynamic time features of the charge-discharge curve, and the spatial distribution features of electrolyte flow anomalies in the spatio-temporally aligned multi-modal data into the fully connected layer to generate corresponding feature encoding vectors respectively.

[0070] It should be noted that when inputting the local temperature gradient features of the infrared thermal image, the dynamic time features of the charge-discharge curve, and the spatial distribution features of electrolyte flow anomalies in the spatio-temporally aligned multi-modal data into the fully connected layer, these features are first converted into a vector form suitable for processing, and then linearly combined through the weight matrix and bias term in the fully connected layer and processed by the activation function to introduce non-linearity, thereby generating corresponding feature encoding vectors.

[0071] Further, each feature type, such as the local temperature gradient feature, is mapped to a high-dimensional space in its original dimension. In this process, the fully connected layer applies a series of transformations to the input features. The weight matrix determines how to combine different feature elements to emphasize or suppress certain specific information, while the bias term provides additional degrees of freedom to adjust the output value, ensuring that the model can learn more complex patterns. Finally, by applying activation functions such as ReLU, the expressive power of the model can be further enhanced, making the generated feature encoding vectors not only contain the key information of the original input features but also enhance the model's ability to distinguish different feature patterns through non-linear transformations, thereby effectively representing each type of feature for use in subsequent steps.

[0072] S2.5: Dynamically allocate the weighted features of multiple modalities based on the feature encoding vectors.

[0073] It should be noted that different types of features (such as the local temperature gradient features of the infrared thermal image, the dynamic time features of the charge-discharge curve, and the spatial distribution features of abnormal electrolyte flow) are converted into feature encoding vectors. These vectors contain the key information of their respective features in specific dimensions.

[0074] Furthermore, the spatio-temporal attention mechanism is used to process these feature encoding vectors. The spatio-temporal attention mechanism can dynamically adjust the importance weights of each feature in both the time and space dimensions. Specifically, in the time dimension, the mechanism evaluates the importance of a feature based on its changing trend over time; in the space dimension, it evaluates based on the performance of the feature at different positions inside the battery. For example, if the voltage change is particularly significant during a certain period, or the temperature gradient in a certain area is extremely high, then these features will be assigned higher weights.

[0075] Specifically, based on these attention scores, a weighted sum is performed on each feature encoding vector to generate a fused feature vector. In this process, important features are given greater weights, while relatively unimportant features are weakened. The purpose of this is to ensure that the finally generated fused feature vector can pay more attention to the information crucial for the assessment of the battery health state, while reducing the influence of irrelevant or redundant features.

[0076] S2.5: Dynamically adjust the weight of the local temperature gradient feature of the infrared thermal image according to the maximum value of the local temperature gradient.

[0077] Furthermore, the weight of the local temperature gradient feature of the infrared thermal image is dynamically adjusted according to the maximum value of the local temperature gradient. First, calculate the temperature gradients of each local area in the image and determine the maximum value among these gradients. Based on this maximum value, adaptively adjust the weight of the temperature gradient feature of each local area. Specifically, if the temperature gradient of a certain local area is close to or equal to the maximum value, it indicates that there may be significant temperature changes in this area, which may be an indication of potential problems (such as hot spots or abnormal temperature rise), so its weight will be correspondingly increased to highlight this key information. On the contrary, if the temperature gradient of a certain area is small, its weight will be relatively reduced to reduce the impact on the overall feature representation. Through this dynamic adjustment mechanism, the model can pay more attention to the areas with important temperature changes, thus more accurately capture the state changes inside the battery and improve the accuracy and reliability of subsequent analysis and prediction.

[0078] S2.6: Dynamically adjust the weight of the dynamic time feature of the charge-discharge curve according to the slope of the charge-discharge curve.

[0079] It should be noted that the weight of the dynamic time feature of the charge-discharge curve is dynamically adjusted according to the slope of the charge-discharge curve. This means that when processing the charge-discharge curve, first calculate the voltage change rate at different time points, that is, the slope of the charge-discharge curve. Then, based on these slope values, adjust the weight of the dynamic time feature at each time point.

[0080] Specifically, if the slope at a certain time point is large, it indicates that the voltage change is significant during this period, which may reflect important events or states in the battery charge-discharge process (such as the rapid charging stage or the high-load discharging stage). Therefore, its weight will be increased accordingly to highlight the information at this critical moment. On the contrary, if the slope at a certain time point is small, its weight will be relatively reduced to reduce the impact on the overall feature representation. Through this dynamic adjustment mechanism, the model can pay more attention to those time periods with significant voltage changes, thus more accurately capturing the key dynamic features in the battery charge-discharge process and improving the accuracy and reliability of subsequent analysis and prediction.

[0081] S2.7: Dynamically adjust the weight of the spatial distribution feature of the abnormal electrolyte flow according to the proportion of the abnormal electrolyte area.

[0082] It should be noted that first calculate the area of each abnormal area in the image and determine the proportion of the area of these abnormal areas in the total area of the entire image. Then, based on this proportion, adjust the weight of the spatial distribution feature of each abnormal area.

[0083] Specifically, if the proportion of the area of a certain abnormal area is large, it indicates that there are significant electrolyte flow problems in this area, which may be an important indication of potential faults or abnormal phenomena. Therefore, its weight will be increased accordingly to highlight the information of this key area. On the contrary, if the proportion of the area of a certain abnormal area is small, its weight will be relatively reduced to reduce the impact on the overall feature representation. Through this dynamic adjustment mechanism, the model can pay more attention to those areas with larger abnormal areas, thus more accurately capturing the state changes of the electrolyte flow inside the battery and improving the accuracy and reliability of subsequent analysis and prediction.

[0084] S2.8: Generate a fused feature vector by concatenating the local temperature gradient feature of the infrared thermal image, the dynamic time feature of the charge-discharge curve, and the spatial distribution feature of the abnormal electrolyte flow in dimension.

[0085] It should be noted that, first of all, each type of feature will be converted into a fixed-length vector representation, and these vectors respectively represent the information of their respective features in specific dimensions. For example, the local temperature gradient feature can be represented as a vector containing the temperature changes in each local area, the dynamic time feature can be represented as a vector containing the voltage change rates at different time points, and the spatial distribution feature can be represented as a vector containing the proportion of the area of each abnormal area. Next, these feature vectors will be concatenated in dimensions, that is, they will be connected together in a certain specified order (such as sorted by feature type or importance) to form a longer single vector.

[0086] Specifically, assume that the dimension of the local temperature gradient feature vector is D1, the dimension of the dynamic time feature vector is D2, and the dimension of the spatial distribution feature vector is D3. Then the total dimension of the concatenated fusion feature vector will be D1 + D2 + D3. This dimension concatenation operation not only retains the original information of each feature, but also enables all features to be uniformly processed in the same vector space, thus providing more comprehensive and integrated input data for subsequent model training and prediction. Finally, the generated fusion feature vector can simultaneously reflect the temperature changes inside the battery, the dynamic behavior during charge and discharge, and the abnormal conditions of the electrolyte flow, laying a solid foundation for accurately evaluating the battery health state.

[0087] S3: Input the fusion feature vector into the GRU-Transformer hybrid model to generate a dynamic capacity prediction value, and identify the aging type by dynamically analyzing historical data.

[0088] S3.1: Input the fusion feature vector into the bidirectional GRU unit in the order of time windows, extract the voltage-temperature-internal resistance time series correlation features under charge and discharge cycles, and output a sequence of hidden state vectors.

[0089] It should be noted that when inputting the fusion feature vector into the bidirectional GRU (bidirectional gated recurrent unit) in the order of time windows, first, the fusion feature vector is segmented according to the set time window size, and the data within each time window represents the comprehensive features of a time segment. Then, the feature vectors within these time windows are sequentially input into the bidirectional GRU unit. Inside the bidirectional GRU unit, the forward and backward GRUs respectively process the time series data: the forward GRU processes the information in the sequence from the past to the future, while the backward GRU processes the same information from the future back to the past. In this way, the bidirectional GRU can obtain all relevant information before and after that moment at each time point, so as to more comprehensively capture the complex time series correlation features among voltage, temperature, and internal resistance under charge and discharge cycles.

[0090] Specifically, within each time step, the bidirectional GRU cell updates its hidden state based on the current input feature vector and the state of the previous time step. This process not only considers information from individual dimensions such as voltage changes, temperature fluctuations, and internal resistance changes but also incorporates the relationships between their interactions. For example, a voltage drop at a certain time point may be related to a previous temperature increase or an increase in internal resistance, and the bidirectional GRU can identify such cross-time-point correlation patterns. As the sequence progresses, the bidirectional GRU gradually extracts and accumulates these temporal correlation features and generates a hidden state vector at each time step, which contains comprehensive information from that time step and all the time steps before and after it.

[0091] S3.2: Input the sequence of hidden state vectors into the Transformer encoder. Through the self-attention mechanism, identify the aging correlation matrix between features at different time steps and generate encoded feature vectors.

[0092] It should be noted that when inputting the sequence of hidden state vectors into the Transformer encoder, first, the self-attention mechanism is used to identify the aging correlation matrix between features at different time steps and generate encoded feature vectors. Specifically, when the sequence of hidden state vectors enters the Transformer encoder, the hidden state vector at each time step is compared and weighted with the hidden state vectors of all other time steps to calculate an aging correlation matrix. This process is achieved through the self-attention mechanism: for each time step, the system calculates a query, key, and value vector, which are obtained through linear transformation from the original hidden state vector. Then, the system performs a dot product operation between the query vector and the key vectors of all time steps and converts it into a probability distribution through the softmax function to determine the importance weights of each time step for other time steps.

[0093] Furthermore, based on these importance weights, a weighted sum of the value vectors at each time step is calculated to obtain a new vector representing the context information of that time step. In this process, the self-attention mechanism can capture complex dependencies between different time steps, especially those patterns related to battery aging. For example, a voltage change at a certain time step may be highly correlated with temperature fluctuations or internal resistance changes at other time steps, and the self-attention mechanism can automatically learn and emphasize these correlations to form the aging correlation matrix.

[0094] S3.3: Based on the fusion of the encoded feature vectors and the state of charge of the real-time battery, calculate the dynamic capacity prediction value through a fully connected layer. The expression is:

[0095]

[0096] Wherein, Q represents the dynamic capacity prediction value, σ represents the Sigmoid activation function, W g represents the gating weight matrix, g represents the gating parameter, C represents the global aging feature vector output by the Transformer encoder, S t represents the battery state parameter vector at the current time t, ⊙ represents element-wise multiplication, represents the weight matrix according to the connection type k and the dimension d of the input feature, d represents the dimension of the input feature, k represents different connection types, F represents the multimodal fusion feature vector, η represents the residual mixing coefficient, Q1 represents the historical capacity mean, W r represents the residual term weight matrix, and r represents the anti-noise compensation term related parameter.

[0097] S3.4: Based on the dynamic capacity prediction value, analyze by extracting the gradient features of the dynamic capacity prediction value and the historical aging pattern library, and identify the aging type.

[0098] It should be noted that the capacity prediction values of the battery at multiple time points are obtained from the dynamic capacity prediction model. Then, calculate the change rate of these capacity prediction values, that is, the gradient features, to capture the trend and rate of the battery capacity changing over time.

[0099] Specifically, extracting the gradient features of the dynamic capacity prediction value involves calculating the capacity difference between adjacent time points and normalizing it to the change rate per unit time. For example, if the battery capacity drops rapidly during a certain period, the gradient feature of that period will show a large negative value, indicating significant aging. On the contrary, if the capacity remains relatively stable, the gradient feature is close to zero.

[0100] Furthermore, compare and analyze these gradient features with the data in the historical aging pattern library. The historical aging pattern library contains typical feature patterns of different aging types, such as cycle aging, calendar aging, or abuse aging, etc. Each aging pattern has its unique capacity decay trend and gradient feature. By comparing the gradient features of the current battery with the patterns in the library, the most matching aging type can be found.

[0101] S4: Based on the aging type, display the lithium-ion distribution through the color change of the electrochromic electrolyte, detect the microscopic impedance using the principle of quantum tunneling, calibrate the battery interface state, and then automatically update the capacity prediction model parameters and output the calibrated capacity value.

[0102] S4.1: Based on the aging type, send a pulse voltage signal proportional to the aging degree to the electrolyte injection unit and activate the color change reaction of the photosensitive compound in the electrolyte.

[0103] Based on the aging type, according to the identified aging type and degree, determine the corresponding pulse voltage signal parameters, which are proportional to the aging degree, to ensure that different severity levels of aging phenomena can be addressed specifically. Then, send the calculated pulse voltage signal to the electrolyte injection unit, which is responsible for applying the pulse voltage signal to the electrolyte inside the battery.

[0104] Furthermore, specific photosensitive compounds are pre-added to the electrolyte, and these compounds are sensitive to electric field changes. When the pulse voltage signal is applied to the electrolyte, the photosensitive compounds will undergo a color change reaction under the action of the electric field. Specifically, the pulse voltage signal will cause the electric field distribution in the electrolyte to change, resulting in a change in the molecular structure of the photosensitive compounds, thus triggering a color change. This color change can directly reflect the state changes in certain regions of the electrolyte (such as lithium-ion deposition or dendrite growth regions). In this way, the microscopic structure changes in the electrolyte can be visually observed and analyzed based on the color change. For example, the color change can indicate the distribution of lithium ions in the electrolyte or show the location and extent of the local impedance increase caused by aging.

[0105] S4.2: Based on the color change reaction of the electrolyte, capture the color intensity distribution map of the lithium-ion deposition area through the multispectral imaging unit and mark the color mutation coordinates;

[0106] It should be noted that first, after the pulse voltage signal is applied, the photosensitive compounds in the electrolyte undergo a color change reaction due to the action of the electric field, and the color changes in different regions reflect the lithium-ion deposition or dendrite growth situation. Next, use the multispectral imaging unit to capture and record these color changes with high resolution. The multispectral imaging unit can simultaneously obtain image data in multiple bands, thereby generating a detailed color intensity distribution map.

[0107] Specifically, the imaging unit scans and images the inside of the battery in different spectral ranges (such as visible light, near-infrared, etc.) through a series of filters or sensor arrays. Each band of the image contains specific color information. By superimposing and analyzing these images, a comprehensive color intensity distribution map can be generated. This map not only shows the color change situation but also can accurately reflect the color intensity differences in each region, thus helping to identify the specific location and extent of lithium-ion deposition.

[0108] Furthermore, during the process of generating the color intensity distribution map, automatic detection and marking of color mutation points in the image are carried out. Specifically, by calculating and comparing the color intensity differences between adjacent pixels, those regions with significant color changes, namely the color mutation coordinates, are identified. For example, if the color intensity of a certain region suddenly increases or decreases, this may indicate the presence of lithium ion deposition or dendrite growth phenomena in that region.

[0109] S4.3: According to the color mutation coordinates, control the movement of the nano-probe array to the target area, apply a constant bias voltage, and start detecting the amplitude change of the current through quantum tunneling current.

[0110] It should be noted that after determining the target area where the nano-probe array needs to move according to the color mutation coordinates, the next operation is to accurately position the probes at these marked coordinates. Once the positioning is completed, a constant bias voltage can be applied between the probe and the target surface. This bias voltage is to prompt electrons to tunnel from one end to the other end, that is, to detect the change of current through the quantum tunneling effect.

[0111] The specific process includes adjusting the distance between the probe and the surface of the lithium dendrite to a known reference spacing to ensure the consistency of conditions during each measurement. Then, apply a specific value of bias voltage between the contact point of the probe and the lithium dendrite, and this voltage value remains unchanged to ensure the stability of the external electric field during the measurement process. Next, start the quantum tunneling current detection mechanism, which usually involves high-sensitivity current detection equipment to capture the tiny current changes caused by the quantum tunneling effect. Since the quantum tunneling current is extremely sensitive to the distance between the probe and the surface, any slight distance change will cause a significant change in the current amplitude.

[0112] S4.4: Based on the amplitude change of the quantum tunneling current and the probe displacement, calculate the local impedance value of the lithium dendrite growth region, and the expression is:

[0113]

[0114] Among them, Z(x,y) represents the local impedance value of the calculated lithium dendrite growth region at the coordinate (x,y), V represents the bias voltage applied by the probe, x represents the horizontal coordinate of the electrolyte region, y represents the vertical coordinate of the electrolyte region, I(x,y) represents the amplitude of the quantum tunneling current at the coordinate (x,y), γ represents the color intensity correction coefficient, G(x,y) represents the photochromic region at (x,y), G represents the color intensity determination threshold, β represents the spacing attenuation coefficient, M represents the reference spacing, and M(x,y) represents the real-time distance between the probe tip and the lithium dendrite surface at (x,y).

[0115] S4.5: Based on the local impedance values, analyze the local impedance changes in the lithium dendrite growth region through electrochemical impedance spectroscopy.

[0116] It should be noted that the process of analyzing the local impedance changes in the lithium dendrite growth region through electrochemical impedance spectroscopy based on the local impedance values is as follows: First, use the local impedance value as a reference point, which reflects the impact of lithium dendrite growth on the local electrical characteristics inside the battery. Then, apply a small-amplitude sinusoidal voltage perturbation to the battery system at different frequencies and measure the corresponding current response.

[0117] Specifically, apply a series of sinusoidal voltage signals ranging from a small frequency (such as millihertz) to a high frequency (such as kilohertz) to the battery through an electrochemical workstation. At each frequency point, record the voltage and current responses of the battery to obtain complex impedance data. These data include real and imaginary part information and can be used to plot a Nyquist plot or a Bode plot to visually observe the change of impedance with frequency.

[0118] By analyzing these impedance data, characteristic changes related to lithium dendrite growth can be identified. For example, in the high-frequency region, the impedance mainly reflects the resistance of the electrolyte and electrode interface; while in the low-frequency region, it more reflects the changes in the diffusion process and charge transfer resistance. The presence of lithium dendrites will change these parameters, resulting in changes in the impedance values in the high-frequency and low-frequency regions. Especially in the low-frequency region, the local impedance increase caused by lithium dendrites may significantly affect the charge transfer process, thus forming obvious characteristics in the impedance spectrum.

[0119] Furthermore, combined with the local impedance values, the results of the electrochemical impedance spectroscopy can be associated with the actual physical location. For example, if the local impedance value of a specific region is high, then higher impedance values and different frequency response characteristics should be observed in the corresponding electrochemical impedance spectrum of that region.

[0120] S4.6: Based on the historical battery charge and discharge data and the interface health status markers, construct a battery interface state evaluation framework, and input the local impedance changes into the evaluation framework to generate corrected interface state parameters.

[0121] It should be noted that the first step in constructing a battery interface state evaluation framework based on historical battery charge and discharge data and interface health state markers is to collect and organize a large amount of relevant historical data. This data includes records of voltage, current, and temperature changes during different charge and discharge cycles of the battery, as well as interface health state markers (such as electrolyte decomposition, lithium deposition, etc.) obtained through experimental or monitoring means. Next, preprocess this data, such as standardization, normalization, and outlier removal, to ensure the consistency and reliability of the data. Then, define a series of characteristic indicators, such as the rate of change of internal resistance, the rate of capacity decay, etc., based on known aging patterns and interface health states, and establish a preliminary evaluation framework. This framework can display the changing trends of different characteristics over time through visualization tools to help understand the evolution law of the battery interface health state.

[0122] Furthermore, input the local impedance change values obtained through electrochemical impedance spectroscopy analysis into the already constructed battery interface state evaluation framework. The algorithms within the framework will conduct a comparative analysis by combining the local impedance changes with the characteristic indicators in the historical data to identify the specific changes in the current battery interface health state. For example, if the local impedance increases significantly, it may indicate the growth of lithium dendrites or other interface problems. Then, based on these analysis results, the framework will adjust and correct the original interface state parameters to generate a new set of interface state parameters that more accurately reflect the current actual health condition of the battery. Finally, these corrected interface state parameters can be used to update the model in the battery management system to provide more accurate state estimation and prediction, thereby optimizing the battery maintenance strategy and extending its service life.

[0123] S4.7: Update the GRU-Transformer hybrid model parameters based on the corrected interface state parameters and output the calibrated capacity value.

[0124] It should be noted that input the corrected interface state parameters (such as internal resistance change, lithium deposition situation, etc.) obtained through the battery interface state evaluation framework into the capacity prediction algorithm. These parameters reflect the more accurate current health condition of the battery; then, the capacity prediction algorithm adjusts its internal model parameters according to these new interface state parameters, such as recalibrating the internal resistance coefficient, updating the aging factor, or adjusting the capacity decay curve, to reflect the latest battery health state; then, recalculate the remaining capacity of the battery using the updated parameters to ensure that the prediction result is closer to the actual situation; finally, after this series of adjustments and calculations, the algorithm outputs a calibrated capacity value, which not only considers the historical operation data of the battery but also incorporates the latest interface health state information, thus providing a more accurate battery capacity estimate.

[0125] This embodiment also provides a computer device, which is applicable to the case of the battery capacity calculation method of the storage and charging system, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the battery capacity calculation method of the storage and charging system proposed in the above embodiment.

[0126] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a button, a trackball, or a touchpad provided on the outer shell of the computer device, or an external keyboard, a touchpad, or a mouse, etc.

[0127] This embodiment also provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the battery capacity calculation method of the storage and charging system proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, abbreviated as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, abbreviated as EPROM), programmable read-only memory (Programmable Red-Only Memory, abbreviated as PROM), read-only memory (Read-Only Memory, abbreviated as ROM), magnetic memory, flash memory, a magnetic disk, or an optical disc.

[0128] In summary, the present invention: ensures the spatio-temporal consistency of data through the timestamp alignment and three-dimensional coordinate conversion algorithm, and solves the problem of single data source in the traditional method; and extracts multi-modal features through technologies such as DSC, BiLSTM, and U-Net, realizing the comprehensive monitoring and dynamic adjustment of the battery health state.

[0129] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and all of them should be covered by the scope of the claims of the present invention.

Claims

1. A method for calculating the battery capacity of a storage and charging system, characterized in that: including Collecting multimodal data of the battery pack, and generating spatiotemporally aligned multimodal data through timestamp alignment and three-dimensional coordinate transformation algorithms Based on the spatiotemporally aligned multimodal data, extracting local temperature gradient features of the infrared thermal image and dynamic time features of the charge-discharge curve through DSC and BiLSTM, segmenting the spatial distribution features of electrolyte flow anomalies in the ultrasonic image using U-Net, and dynamically allocating weights for multimodal features through a spatiotemporal attention mechanism to generate a fused feature vector Inputting the fused feature vector into a GRU-Transformer hybrid model to generate a dynamic capacity prediction value, and identifying the aging type by dynamically analyzing historical data Based on the aging type, displaying the lithium-ion distribution through the color change of the photochromic electrolyte, calculating the local impedance value of the lithium dendrite growth region using the principle of quantum tunneling, calibrating the battery interface state, automatically updating the parameters of the GRU-Transformer hybrid model, and outputting the calibrated capacity value 2. The method for calculating the battery capacity of the storage and charging system according to claim 1, wherein: The multimodal data includes infrared thermal imaging temperature field data, ultrasonic electrolyte distribution data, visible light appearance deformation image data, and electrochemical parameter data 3. The battery capacity calculation method of the storage and charging system according to claim 2, wherein: The method for generating spatiotemporally aligned multimodal data through timestamp alignment and three-dimensional coordinate transformation algorithms is as follows Adding synchronous timestamps to the multimodal data, and compensating for the sampling rate difference through a cubic spline interpolation algorithm to generate time-synchronized data values Based on the three-dimensional structure model of the battery pack, mapping the time-synchronized data values to the same coordinate system, eliminating the position deviation through a rotation and translation matrix, and outputting the spatiotemporally aligned multimodal data 4. The battery capacity calculation method of the storage and charging system according to claim 3, wherein: The method for extracting local temperature gradient features of the infrared thermal image and dynamic time features of the charge-discharge curve through DSC and BiLSTM, and segmenting the spatial distribution features of electrolyte flow anomalies in the ultrasonic image using U-Net based on the spatiotemporally aligned multimodal data is as follows Inputting the infrared thermal image in the spatiotemporally aligned multimodal data into DSC, and extracting local temperature gradient features layer by layer through convolutional kernels Inputting the charge-discharge voltage curve in the spatiotemporally aligned multimodal data into BiLSTM, and extracting the dynamic time features of the charge-discharge curve through the bidirectional sequence processing ability of BiLSTM Inputting the ultrasonic image in the spatiotemporally aligned multimodal data into the U-Net segmentation network, and generating the spatial distribution features of electrolyte flow anomalies through an encoder-decoder structure 5. The method for calculating the battery capacity of the storage and charging system according to claim 4, characterized in that: The method for dynamically allocating weights for multimodal features through a spatiotemporal attention mechanism to generate a fused feature vector is as follows Inputting the local temperature gradient features of the infrared thermal image, the dynamic time features of the charge-discharge curve, and the spatial distribution features of electrolyte flow anomalies in the spatiotemporally aligned multimodal data into a fully connected layer to generate corresponding feature encoding vectors respectively Based on the feature encoding vectors, dynamically allocating weight features for multimodals through a spatiotemporal attention mechanism Dynamically adjusting the weight of the local temperature gradient features of the infrared thermal image according to the maximum value of the local temperature gradient Dynamically adjusting the weight of the dynamic time features of the charge-discharge curve according to the slope of the charge-discharge curve Dynamically adjust the weight of the spatial distribution characteristics of abnormal electrolyte flow according to the proportion of the area of the abnormal electrolyte region; Generate a fused feature vector by splicing the local temperature gradient characteristics of the infrared thermal image, the dynamic time characteristics of the charge-discharge curve, and the spatial distribution characteristics of abnormal electrolyte flow according to dimensions.

6. The method for calculating the battery capacity of the storage and charging system according to claim 5, characterized in that: Input the fused feature vector into the GRU-Transformer hybrid model, output the dynamic capacity prediction value after fusion, and identify the aging type through dynamic analysis of historical data. The specific steps are as follows: Input the fused feature vector into the bidirectional GRU unit in the order of time windows, extract the voltage-temperature-internal resistance time series correlation characteristics under charge-discharge cycles, and output a sequence of hidden state vectors; Input the sequence of hidden state vectors into the Transformer encoder, identify the aging correlation matrix between features at different time steps through the self-attention mechanism, and generate the encoded feature vector; Fuse the encoded feature vector with the state of charge of the real-time battery, and calculate the dynamic capacity prediction value through the fully connected layer; Based on the dynamic capacity prediction value, analyze by extracting the gradient characteristics of the dynamic capacity prediction value and the historical aging pattern library, and identify the aging type.

7. The method for calculating the battery capacity of the storage and charging system according to claim 6, characterized in that: Based on the aging type, display the lithium ion distribution through the color change of the electrochromic electrolyte, and calculate the local impedance value of the lithium dendrite growth region using the principle of quantum tunneling. The specific steps are as follows: Based on the aging type, send a pulsed voltage signal proportional to the aging degree to the electrolyte injection unit, and activate the color change reaction of the photosensitive compound in the electrolyte; Based on the color change reaction of the electrolyte, capture the color intensity distribution map of the lithium ion deposition region through the multispectral imaging unit, and mark the coordinates of color mutations; According to the coordinates of color mutations, control the nano-probe array to move to the target area, apply a constant bias voltage and start detecting the amplitude change of the current of the quantum tunneling current; Calculate the local impedance value of the lithium dendrite growth region based on the amplitude change of the quantum tunneling current and the probe displacement.

8. The method for calculating the battery capacity of the storage and charging system according to claim 7, characterized in that: Automatically update the capacity prediction model parameters after calibrating the battery interface state, and output the calibrated capacity value. The specific steps are as follows: Based on the local impedance value, analyze the local impedance change of the lithium dendrite growth region through electrochemical impedance spectroscopy; Based on the historical battery charge-discharge data and the interface health state mark, construct a battery interface state evaluation framework, and input the local impedance change into the evaluation framework to generate the corrected interface state parameters; Based on the corrected interface state parameters, update the parameters of the GRU-Transformer hybrid model, and output the calibrated capacity value.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the battery capacity calculation method of the storage and charging system according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the battery capacity calculation method of the storage and charging system according to any one of claims 1 to 8.

Images